采用梯度平滑措施的自适应符号算法

以自适应回波对消为对象 ,讨论了两种自适应符号算法的梯度估计平滑方法。从理论上对其中一种平滑方案导出的自适应算法的性能进行了详尽的分析。在输入为联合高斯的情况下 ,给出了权值一阶矩和二阶矩的收敛特性公式。结果表明 ,与符号算法相比 ,平滑算法的收敛速度与符号算法基本一致 ,但稳态误差更小 ,自适应步长的取值范围更大。

On the Gradient Smoothed Processing of Adaptive Sign Algorithm

Two smoothed gradient estimate methods used in adaptive sign algorithm applied to adaptive echo canceling have been discussed in this paper. One of the resulting algorithm performance is analyzed in detail including the mean and the second moment convergence behavior of weights under the condition that input signal is zero-mean finite variance jointly Gauss process. It indicates that the scheme can enlarge the allowable adaptive step size and improve the steady state error without affecting the convergence speed.